Well, it's the policy and the structure of the organization. It's been the algorithm that we've spent basically Victorian area for and the bureaucratic world forward trying to deal with. So now we may be putting it into the to the machine loop and it's going to be the same fears and frustrations all over again. So anyway, let me go to my next question that I had on that. And we just why the risk might be fundamentally different a project like this in your experience, guys, what is the smallest application that you have seen where someone has successfully rolled it out?

I'll talk about some of the small ones. So if you need any time that you're going to keep the human in the loop too, it can be a little bit smaller. It doesn't need to be as accurate, right? We talked about autonomous where the goal of that is to remove the human from the loop. That needs to be pretty accurate. I mean, you think about a car, right? If you're going to fully remove me from a bit scary, you better be darn near accurate. We've seen recommendation systems. So that can be small and you're just recommending different products. You can implement that fairly quickly. We've also seen some image recognition where it's just identifying small things in an image. Also a lot of automation style, which automation, we always say is not intelligence, but we've seen a lot of people do automation and that can be really fast, especially how that quick ROI return on investment. And then another thing that we've talked about is when we always say think big but start small. So if you are building a chatbot, for example, you're going to use natural language processing, think about the question that gets asked the most and how you can relieve maybe call center volume or just volume on your site and address that specific question. So the US Postal Service put together a chatbot and they said the biggest question that they get asked is tracking a package. So that's what the bot answered. Not any of the other questions, just that. So start small, then you can really see a

return.